Akademik Veri Yönetim Sistemi
Journal of Stomatology, cilt.78, sa.1, ss.42-51, 2025 (Scopus)
INTRODUCTION: Tooth detection and numbering is a fundamental issue that has been the subject of research projects applying artificial intelligence in dental radiography evaluation. OBJECTIVES: The objectives of the current study were to build a deep learning model for detecting and numbering teeth, and to evaluate the model’s performance. MATERIAL AND METHODS: Retrospective data collection was done from 3,491 bitewing radiographs of randomly selected individuals. Confusion matrix was used to calculate sensitivity, precision, and true positive and false positive/negative values to examine the performance of the algorithm. RESULTS: The sensitivity and precision were 0.9940 and 1 for the classifying task, respectively. In addition, the predicted F1 score was 0.99970, demonstrating a favorable balance between recall and precision. On the right side, the IDF1 score was 89%, with a confidence level of 0.73. The mAP for all classes was high, accurately modeling 90.9% detections with 0.5 threshold. On the left side, the IDF1 score was 89%, with a confidence level of 0.37. The mAP for all classes was high, accurately modeling 91% detections with 0.5 threshold. CONCLUSIONS: The most disadvantageous feature of the bitewing radiograph is that sometimes various areas of the tooth cannot be completely observed. This situation leads to a reduction in the detection ability of the model. However, this research shows that convolutional neural networks algorithms can be very accurate and effective in detecting and numbering of the teeth.